Time series data missing value automatic filling method, system and device

Abstract

The invention provides a time series data missing value automatic filling method, system and device. Through a random mask and a neural network decomposition technology, the time sequence data is decomposed into superposition of different modes, and automatic feature extraction is realized. Therefore, a more accurate and effective end-to-end time series data missing value filling method is constructed. The method comprises three steps of data preparation, model training and model use. In the data preparation step, original time series data is acquired for data preprocessing, a random mask is constructed according to a given missing rate, and the newly generated random mask and the corresponding original data are used as a new data set. In the model training step, the new data set generatedin the data preparation step is used for model training so as to construct a model based on neural network decomposition. In the model using step, corresponding masks are constructed for the time series data with missing values, and the trained model is used for filling the missing values of the time series data.

Description

technical field [0001] The invention relates to the technical field of machine learning, in particular to a method, system and equipment for automatically filling missing values ​​of time series data. Background technique [0002] With the development of deep learning and neural networks, the analysis and processing of time series data has attracted more and more attention. Such as meteorology, medical care, transportation, water affairs, etc. However, these actual time series data will inevitably produce missing values ​​due to various reasons. In order to better analyze and utilize these data, missing value processing and filling must be performed first. Missing values ​​of time series data often have nonlinear and dynamic correlations with other values. Traditional filling methods such as zero filling, mean value supplementation and EM algorithm cannot effectively deal with this nonlinear and dynamic correlation. However, some methods based on LSTM model regard missing ...

AI Technical Summary

Problems solved by technology

On the other hand, the time series data itself often has some direct or hidden patterns, and the previous methods often ignore these patterns, and of course cannot use these patterns to obtain more accurate and effective missing value filling effects

Images